Rethinking the quest for provenance.

Budd, P. ; Haggerty, R. ; Pollard, A.M. 等

One of the larger - and more expensive - present programmes of study in archaeological science explores the provenance of prehistoric bronzes from the Mediterranean. What are the bases of research? What will the findings tell us about the real place of metal as it moved in the ancient world?

The provenance postulate

The availability of spectrographic methods in the 1930s made it possible to analyse large numbers of ancient metal artefacts with a view determining their provenance. Some of these studies (e.g. Pittioni 1957) followed from a 19th-century tradition, pioneered by the Austrian scholar Wocel (see Caley 1951; 1967), in which it was proposed that the impurities in ancient copper artefacts would directly reflect those in the ores from which they were smelted. Some scientifically informed commentators (Thompson 1958) expressed concern as to whether this naive insistence was justified, but the work continued in the absence of any detailed understanding either of the geology of metal ores or of the chemistry of their smelting. Although most researchers have come to recognize the complexity and limitations of compositional data (see for example Pernicka 1995), Pittioni-style provenance studies continued for decades. Despite Thompson's warning, tens of thousands of prehistoric metal artefacts were subsequently analysed. The Studien zu den Anfangen der Metallurgie (SAM) analytical programme (Junghans et al. 1960; 1968; 1974) and the huge analytical programme in the former Soviet Union (Chernykh 1994) stand as the two largest monuments to this endeavour. Interpretation of the data in these compilations has generally used statistical procedures to extract groups of artefacts having similar composition. On the broadest level, such groupings are undoubtedly significant. In much of Eurasia, the use of unalloyed copper pre-dates the use of arsenic-rich material, and this is later succeeded by tin-bronze. On a more detailed level attempts to link compositionally similar artefacts to common geographical sources have almost always proved inconclusive. Artefacts grouped on the basis of composition have often come from unrelated contexts, sometimes widely scattered across vast areas, as Butler & Van der Waals (1964) noticed in commenting on the SAM programme.

When the assumptions which underpin these approaches are examined, it is easy to see the pitfalls. For trace element provenancing to succeed it is necessary to assume that the artefacts under consideration were fabricated using similar manufacturing processes, derived from a strictly limited number of sources and smelted in such a way as to produce a metallic product with a limited range of impurities. Furthermore, one is obliged to make further fundamental assumptions about alloying and recycling. Whereas some copper deposits may contain distinctive traces of particular trace or minor elements, very few have been studied mineralogically to the point where quantitative estimates of different mineral species' contributions to ore 'as-mined in prehistory' can be made. Similarly, there is little detailed understanding of the behaviour of impurities in primitive smelting processes. Simple thermodynamic models are problematic and many experimental studies have been little more than imaginative reconstructions of hypothetical processes, sometimes using inappropriate materials and poor control and monitoring.

Under these circumstances modern researchers are rightly circumspect in their use of impurity data for provenance studies, although useful studies have been undertaken for small regions; attempting for example to outline compositional variation within artefacts already grouped archaeologically by virtue of typology or context (e.g. Begemann et al. 1995). It is now clear that different deposits may share closely similar geochemical characteristics so that particular copper deposits simply will not yield metal of unique minor or trace element composition. On the other hand, some deposits may be distinctive from some others with respect to certain elements, and Pernicka (1995) argues cogently that trace element analysis offers the potential to aide the discrimination of ore sources in some cases.

These observations about the studies of a generation ago and the limitations of trace element analysis are paralleled in recent work on lead isotopes - work regarded as the promising successor to elemental analysis for provenance study. Some practitioners of lead isotope analysis were amongst the most vociferous critics of trace element metal provenancing (e.g. Gale & Stos-Gale 1982); yet, in developing their methods, these workers themselves retain some of the same assumptions regarding the composition of ore bodies and concerning the nature of metal production and use in prehistoric societies. Although not obvious in the early years of the large-scale lead isotope programmes, when the numbers of ore bodies and artefacts analysed were relatively small, as the data have accumulated so too have many familiar problems. These difficulties could undermine the whole lead isotope provenancing technique if they lead to a loss of faith of the sort that overtook the SAM programme (Coles 1982). In our opinion, this need not be the case. Lead isotope studies can be helpful and relevant in archaeology. We believe however, that more realistic archaeological interpretations of the analytical data are required; a more detailed examination of archaeological lead isotope methodology and the deconstruction of aspects of the current framework are essential to this.

Deconstructing lead isotope provenancing

A vigorous debate on the interpretation of lead isotope data has recently taken place in Archaeometry (Sayre et al. 1992a; 1992b; 1993; Gale & Stos-Gale 1992; 1993; Leese 1992; Pernicka 1992; 1993; Reedy & Reedy 1992; Budd et al. 1993a; 1993b) and in the Journal of Mediterranean Archaeology where it focuses on the Aegean Bronze Age (Budd et al. 1995a; 1995b; Gale & Stos-Gale 1995; Pernicka 1995; Sayre et al. 1995; Hall 1995; Muhly 1995). However, this concern over lead isotope research is not new. Some archaeologists have, for a number of years, been drawing attention to a growing gulf between the conclusions drawn from some lead isotope programmes, and those of their own research work (e.g. Cherry & Knapp 1991; Knapp in press; Knapp et al. 1988; Muhly 1991; 1995). This controversy has disconcerted those who would like to rely on the analytical results, frustration evident in a recent editorial in this journal (Chippindale 1994). It is becoming clear that the interpretation of lead isotope data has not taken place within a framework which reflects the true complexity either of ore deposits, or - perhaps more importantly - of metal supply and circulation in the ancient world. We contend that this crisis of confidence stems, not from lead isotope measurements themselves, but from their interpretation.

As modern mass spectrometers are capable of measuring lead isotope ratios to a high degree of precision, one should be able to have considerable confidence in the accuracy of published data. The preparation and analysis of lead-bearing metals or minerals is a routine and simple task usually undertaken by experienced technicians. Lead-poor materials (Cypriot copper ores, for example) can be more problematic, but again tried and tested methods have been developed to process samples on a routine basis. Unfortunately, recent experience with re-measuring 'outliers' (Gale & Stos-Gale 1992) and the rejection of old measurements in new data compilations (Stos-Gale et al. 1995) have shown that some measurements appear not to have been of the highest precision. This makes it impossible to define accurately the full extent of the Cypriot and other important ore source fields, a situation that could be resolved by a systematic re-measurement programme.

The relative simplicity of lead isotope analysis extends to the interpretation of lead isotope data. As there are only four naturally occurring lead isotopes which must add up to 100%, there are only three independent variables. Every sample can be completely characterized by three numbers and all that is required to express every attribute of the data is a three-dimensional plot. Multivariate analyses (Sayre et al. 1992) or discriminant function analyses (Gale 1991) can be applied, but the usefulness of such approaches is strictly limited. With only three variables and three axes on which to plot them, sophisticated multivariate approaches are, at best, unnecessary. Discriminant function analyses can be used to select rapidly the best projection of three-dimensional data in order to see the greatest separation between groups of samples from different ore deposits. However, it can only be used to calculate the probability of an unknown sample belonging to one such group under the right circumstances: where all possible groups are represented, and where all are fully characterized by statistically significant numbers of samples. This is a rare happening in archaeometallurgy.

All that is actually required to interpret lead isotope data is a three-dimensional plot of the ore data and artefacts under consideration, together with appropriate computer software to rotate and view it. This can be done using the humblest personal computer. There is no reason why archaeologists or archaeological scientists should be marginalized in the interpretation of lead isotope data as even the most traditionally based archaeologist routinely deals with much more complex data-sets. What is needed, in the words of one of the referees of this paper, is 'publication of the mineral data - for samples of stated archaeological and geological characteristics - in a simple, straightforward and workable way'.

Proponents of the lead isotope method of provenancing initially hoped that specific metalliferous regions, be they individual islands such as Cyprus or Kythnos or mainland deposits such as Lavrion, would have their characteristic isotope 'signatures' - each significantly different from one another. In this sense, the programmes can be seen as the isotopic analogue of those of SAM or Pittioni. There would be some natural variation in lead isotope composition within each region, but initial measurements seemed to justify optimism that this would be relatively small. This being so, each region could be characterized by a relatively tight grouping of similar ratios which could be plotted as a 'field' on a three-dimensional diagram or on pairs of two-dimensional plots. The variation in the four naturally occurring lead isotopes in an ore body is related to the genesis and age of the deposit. For archaeological purposes this detailed geological information is largely superfluous. Provided that each ore field is characterized by sufficient numbers of analyses and the variation between sources in the region under consideration remains large, then it should be possible to comment on the provenance of objects of unknown origin.

It was never going to be practical to define every ore deposit in the ancient world, and workers were initially forced to define source fields on less than ideal numbers of analyses. Provided that the analysts were justified in their assumptions, regarding the limited spread of data from an individual ore body and the variation between fields within the region under scrutiny, then provenance analysis could proceed with costs kept within reasonable bounds. Unfortunately, it is now clear that their initial assumptions about ore bodies were not well founded. As the work progressed, it became evident that the variation within many of the ore sources of interest was much greater than had been hoped and that many fields defined in a region, such as the Mediterranean, overlapped each other.

The often-heated debate in Archaeometry and JMA essentially centres on this point, in a detailed discussion likely to lie beyond the interest of any but the most scientifically minded archaeologists. At the heart of the argument is this: do ore source fields, defined using statistical procedures on selected data rather than the required number of analytical measurements, have any validity? We maintain that they do not. In order to maintain apparent separation between source fields, the Oxford group have suggested that all of the ore sources of the Mediterranean have lead isotope fields which display little variation and therefore form very tight and discrete fields on plots of the data. They imply, incorrectly, that this is because all of the ore deposits in question conform to a model based on a single emplacement event in geological time. In reality only a small minority of ore deposits meet the stringent requirements for this type of geological conformability. One of the most important deposits, Cyprus, is one such rare example, but other Mediterranean deposits certainly are not.

The Brookhaven/Smithsonian group have taken this questionable procedure a stage further. Ignoring the fundamental geology that underpins lead isotope analysis, they have assumed that data from each ore field would be normally distributed with respect to each of the three measured ratios (Sayre et al. 1992a). This assumption of tri-variate normality is not likely to be correct (Cherry & Knapp 1991; Budd et al. 1993a; 1993b; Scaife et al. in press). The source fields that seem more precisely defined are achieved by rejecting 'outlying' samples collected from an ore body but somehow not thought to be representative of its lead isotope field (see Budd et al. 1993a). In attempts to define source fields still more tightly, some workers have proposed breaking up some, already ill-defined, source fields into subsets (Sayre et al. 1992; Gale & Stos-Gale 1992). This can only lead to confusion. In extreme cases proposed sub-groups have been defined by as few as two measurements and represent single mines whereas neighbouring mines, only a few kilometres distant, are supposedly represented by entirely different fields!

A full consideration of the geology of ore formation suggests that compact and discrete lead isotope fields for individual ore deposits are likely to be the exception rather than the rule. In contrast to trace element measurements, there are reasons to expect that some deposits (those 'conformable' deposits which more or less correspond to single-age models of emplacement) will have a narrow range of isotopic values; but this can only be discerned by systematic sampling. Characterizing deposits by terming some measurements 'outliers' and then igoring them is simply bad science. So it is that the failings of the lead isotope methodologies proposed by Sayre et al. (1992) and the sub-fields suggested by the Oxford group in response (Gale & Stos-Gale 1992) are the same as those which afflicted the SAM study. The groupings - trace element or isotopic - created by this form of analysis are essentially statistical artefacts which may or may not coincide with entities that have geological or archaeological integrity. The SAM group (incorrectly) assumed that artefacts which were statistically associated in terms of their impurity patterns must derive from the same source, and that artefacts falling outside the group must be derived from a different source. The Brookhaven/Smithsonian group have (incorrectly) assumed that ores (and slags) which are statistically associated in terms of their lead isotope composition must derive from the same source, and that those falling outside the statistical group must be derived from a different source.

With the collapse of the simplistic provenancing approach, the challenge for researchers today is to develop a framework for the archaeological interpretation of lead isotope data in which the limitations imposed by geology or by the complexity of metal production and use are acknowledged and freely discussed. This discussion must not be the exclusive preserve of a restricted group of archaeological scientists, for the answers,involve integrating archaeological information to the greatest possible extent. Kept apart, the isotope data will never yield meaningful patterns.

A new framework for metals analysis

The question of how ore deposits can be defined in geological or chemical terms begs a profound question: 'what constitutes an ore source for prehistoric metallurgy?' So far, ore sources have been considered in terms of islands or large mainland deposits, but this concept is breaking down. The attempts by the Brookhaven/Smithsonian and Oxford groups to sub-divide the Cyprus field into component parts has been discussed (see above and Budd et al. 1995). Unfortunately, the relationship between the lead isotope ratio of an ore deposit and its geographical location is complex. It does not necessarily follow that deposits which are close together will have similar compositions. We believe it is not useful to attempt to characterize individual mines within a metalliferous region such as Cyprus for archaeological as well as geological reasons. Current data show that individual Cypriot mines have ranges of lead isotope ratios which overlap one another (Budd et al. 1995a), but even if they could be accurately resolved there are further objections. There are specific localities at which copper mineralization would have been accessible to ancient miners, but the number of potential sources is vast. Outcrops insignificant in modern economic terms could have yielded plentiful resources for ancient production. Ancient mining localities may have been destroyed by subsequent activity or worked to the extent that representative minerals are no longer available to be sampled. It is impossible to know whether one has measured all the possible ore sources in a region and highly likely that one has not.

The whole concept of metal artefacts being derived from a single raw material source is problematic in the complex craft and trading world of the Late Bronze Age Aegean in which metals are accumulated in bulk and transported long distances. Such ideas are a relic of analogies with modern industries based on particular raw materials, utilising a specialist work force and specific technology. These views are now being independently rejected as a model for prehistoric metal-working (Sherratt 1994; Budd & Taylor 1995; Budd et al. in press). We believe that it is more reasonable to see oxhide ingots as part of a body of smelted copper which might well have been made from raw materials from more than one source. This idea, discussed at length elsewhere (Budd et al. 1995a; 1995b), can explain the relatively tight range of lead isotope compositions of the ingots and their distribution on bi-variate plots, centring towards one side of the Cypriot field. In any system involving the random mixing of primary smelting products from different sources, any degree of weighting (such that one source tends to introduce more copper and another less and so on), will result in secondary material which has a smaller range of lead isotope ratios than any of the individual source groups. If significant recycling is then involved in artefact production, the cluster of values will be tighter still. The position of any such cluster will be somewhere between those of the various sources, depending on their relative contributions. The problem is that it will not be possible to know where all the sources may have been, or even if they are all represented by the available measurements. Extracting the component parts from such a system might be very difficult.

Clearly, as happened with trace element analysis, the application of lead isotopes to provenancing archaeological copper alloys is becoming a proposition far more complex than was initially anticipated. Where the limits of variation of a given ore body are fully defined, any artefact which does not fit within the bounds of the field and its attendant experimental errors is known to derive, at least in part, from material from elsewhere. The inescapable difficulty is that an artefact that falls outside an ore source field is not necessarily wholly derived from another ore. If one has independent evidence that the group of artefacts under investigation were produced from a single source, then it may be possible to draw conclusions regarding locations from which they did not come. In more complex - and perhaps more usual - situations, large quantities of copper would have been in circulation over areas with more than one raw material source; then it may never be realistic to expect lead isotope or trace element measurements to assign provenance unambiguously. Analytical data might be able only to exclude sources which are highly unlikely to have contributed significantly to the metalwork under investigation. Perhaps the whole concept of provenance is redundant as far as prehistoric metals are concerned. Why should we expect, in the complex later prehistoric world, that each object was made from copper extracted only from a single locality and that, once produced, copper alloys were never re-used? The complicated fabrication history and widely scattered iconographic influences attested in spectacular later prehistoric metalwork like Denmark's famous Gundestrup cauldron (Taylor 1992) argue persuasively against that narrow view.

There are ways forward. The Heidelberg/ Mainz group have explored the combination of lead isotope with trace element data to achieve better discrimination between raw material sources, but have always accepted that this would not work if recycling and metal mixing were commonplace (Pernicka et al. 1984). There is some hope that other isotopic systems may indicate the extent of recycling (Budd et al. 1995c). Perhaps the long-term answer, however, is to change the question. The oxhide ingots of copper that now look less attractive for provenance studies remain just as important to the study of Bronze Age production and trade (Knapp in press). This observation is underlain by a crucial point: provenancing is not the only way in which to study the organization of prehistoric metal production and use. Perhaps detecting change in the pattern of metal procurement and use is more useful than assigning provenance. Knapp (in press) has observed that some 30% of the analysed metal finds from a large number of Late Bronze Age Cypriot sites have lead isotope ratios inconsistent with the Cyprus field and suggests that mixing and re-melting of metal - for which there is also evidence from coeval 'foundry' hoards - may have played a role in this. Putting this together with the proposal that some oxhide ingots may have been made from copper originating from more than one location, he goes on to speculate over the possible implications for the development of Late Bronze Age social organization. In this view recycling and the pooling of copper from multiple sources is seen not simply as a back-projection from our contemporary ideas, but as genuinely represented in the material record, suggesting social interactions across cultural borders in the development of a regional economic sphere. Late Bronze Age metallurgy is seen as a complex system in which multiple sources feed into a koine of stock material circulating by virtue of trade, through exchange and perhaps inferring the intervention of developing regional political and/or magico-religious authorities.

As with the large-scale trace-element analytical programmes that preceded it, lead isotope analysis has brought us some way towards unlocking the question of provenance, but definitive answers elude us. Lead isotope ratios will, and should, continue to be measured as part of archaeometallurgical research programmes, but a more flexible approach to their interpretation is required. In particular there is a need to re-evaluate the emphasis on provenance and to develop a framework better suited to the analytical tools we have. This requires a clear dialogue between analysts and the field-workers they support in an environment of rigorous scientific review of approaches and methodologies. These changes are likely to come about naturally as the days become numbered for 'large equipment' facilities for the sole support of archaeology, but scientists must not be allowed to develop an arrogant monopoly in the interpretation of analytical data. Archaeologists must be directly involved in this and in the formulation of relevant but testable hypotheses which move beyond the tired old idea of provenance.

A reliable base of data from lead isotope determinations

In a welcome development since this paper was first drafted, the editor of the journal Archaeometry has undertaken to provide a rapid method of publication for archaeological lead-isotope data in a similar form to the, now familiar, radiocarbon date lists. We feel sure that this will help to clear the current backlog of unpublished measurements and hope that it will result in a comprehensive and reliable database to facilitate the wider interpretation for which we have argued.

The Archaeometry database is intended, for the deposits covered, to be a definitive statement of what is regarded as current reliable data from the Isotrace Laboratory in Oxford (Gale pers. comm.). The measurements tabulated in the first publication in the series - on ores from the western Mediterranean (Stos-Gale et al. 1995) - are a mixture of new and previously published analyses. Some previously published samples are omitted; some because they are only single measurements from otherwise uncharacterized deposits for which more data are expected, and some because they have been rejected until re-measured on grounds of precision, possible contamination and instrument problems (Gale pers. comm.). The specific criteria for data selection are not always made clear in this initial paper, but publication of an explanation of the guidelines underlying data selection and of more, recently acquired, western Mediterranean ore data is anticipated in the near future (Gale pers. comm.).

We strongly support this effort to produce a straightforward and reliable database of lead isotope measurements for ore deposits of archaeological interest. Clearly, it is advantageous to remove bad or suspect data in the construction of ore fields and we are pleased to note that a comprehensive re-measurement programme is now under way at Oxford with this objective. There is, however, a danger that the temporary omission of data from some deposits, pending further or repeat measurements, will lead to a false impression of the overall spread of data from a region; suggesting a more limited field for the region than is actually the case. We note, for instance, that one immediate effect of the omissions from the new Sardinian database is to create the impression of smaller, more discrete, field with only a few outlying samples. We are confident, however, that any danger of misinterpretation can be avoided by comprehensive sampling and analysis and we look forward to the further publications which are promised.

Acknowledgements. Two of the authors (PB and BS) are supported by the Natural Environment Research Council (UK) and one (RGT) by the Australian Research Council. Many thanks to Tim Taylor and referees for many helpful comments and suggestions on earlier versions of this manuscript.

References

BEGEMANN, F., E. PERNICKA & S. SCHMITT-STRECKER. 1995. Thermi on Lesbos: a case study of changing trade patterns, Oxford Journal of Archaeology 14(2): 123-36.

BUDD, P., D. GALE, A.M. POLLARD, R.G. THOMAS & P.A. WILLIAMS. 1993a. Evaluating lead isotope data: further observations, Archaeometry 35(2): 241-7.

1993b. Evaluating lead isotope data: further observations - reply, Archaeometry 35(2): 262-3.

BUDD, P., A.M. POLLARD, B. SCAIFE & R.G. THOMAS. 1995a. Ox-hide ingots, recycling and the Mediterranean metals trade, Journal of Mediterranean Archaeology 8(1): 1-32.

1995b. Lead isotopes and oxhide ingots: a final comment, Journal of Mediterranean Archaeology 8(1): 70-75.

BUDD, P., R. HAGGERTY, A.M. POLLARD, B. SCAIFE & R.G. THOMAS. 1995c. New heavy isotope studies in archaeology, Israel Journal of Chemistry 35(2): 125-30.

BUDD, P., B. SCAIFE, T. TAYLOR & R.G. THOMAS. In press. Untangling the web: some new views on the origins of prehistoric metallurgy, Historical Metallurgy.

BUDD, P. & T. TAYLOR. 1995. The faerie smith versus the bronze industry: magic versus science in the interpretation of prehistoric metal-making, World Archaeology 27(1): 133-43.

BUTLER, J.J. & J.D. VAN DER WAALS. 1964. Metals Analysis, SAM 1 and European prehistory, Helinium 4: 3-39.

CALEY, E.R. 1951. Early history and literature of archaeological chemistry, Journal of Chemical Education 28: 64-6.

1967. The early history of chemistry in the service of archaeology, Journal of Chemical Education 44(3): 120-23.

CHERNYKH, E.N. 1994. Ancient metallurgy in the USSR. Cambridge. Cambridge University Press.

CHERRY, J.F. & A.B. KNAPP. 1991. Quantitative provenance studies and Bronze Age trade in the Mediterranean: some preliminary reflections, in Gale (ed.): 90-119.

CHIPPINDALE, C. 1994. Editorial, Antiquity 68: 1-9.

COLES, J.M. 1982. The Bronze Age in Northwestern Europe: problems and advances. in F. Wendorf & A.E. Close (ed.), Advances in World Archaeology: 1, 265-321. New York (NY): Academic Press.

GALE, N.H. 1991a. Copper oxhide ingots: their origin and their place in the Bronze Age metals trade in the Mediterranean, in Gale (ed.): 197-239.

(ed.). 1991b. Bronze Age trade in the Mediterranean: 90-119. Jonsered: Paul Astroms Forlag. Studies in Mediterranean Archaeology 90.

GALE, N.H. & Z.A. STOS-GALE. 1982. Bronze Age copper sources in the Mediterranean: a new approach, Science 216(4541): 11-19.

1992. Evaluating lead isotope data [comments on Sayre et al. 1992a], Archaeometry 34(2): 311-17.

1993. Comments [on Budd et al. 1993a], Archaeometry 35(2): 252-9.

1995. Comments [on Budd et al. 1995a], Journal of Mediterranean Archaeology 8(1): 33-41.

HALL, M. 1995. Comments [Budd et al. 1995a], Journal of Mediterranean Archaeology 8(1): 42-4.

JUNGHANS, S., E. SANGMEISTER & M. SCHRODER. 1960. Metallanalysen kupferzeitlicher und frubronzezeitlicher Bodenfunde aus Europa. Berlin: Gebr. Mann.

1968. Kupfer und Bronze in der fruhen Metallzeit Europas. Katalog der Analysen Nr: 985-10040. Berlin: Gebr. Mann.

1974. Kupfer und Bronze in der fruhen Metallzeit Europas. Berlin: Gebr. Mann.

KNAPP, A.B. In press. Provenance studies and the Bronze Age Mediterranean: an archaeological perspective. Proceedings of a conference on science and archaeology: towards an interdisciplinary approach to studying the past, Harvard University, Cambridge (MA), October 1994.

KNAPP, A.B., J.D. MUHLY & P.M. MUHLY. 1988. To hoard is human: the metal deposits of LC IIC-LC III, Report of the Department of Antiquities, Cyprus: 233-62.

LEESE, M.N. 1992. Evaluating lead isotope data [comments on Sayre et al. 1992a], Archaeometry 34(2): 318-22.

MUHLY, J.D. 1991. The development of copper metallurgy in Late Bronze Age Cyprus, in Gale (ed.): 180-96. Jonsered: Paul Astroms Forlag. Studies in Mediterranean Archaeology 90.

1995. Lead isotope analysis and the archaeologist [comments Budd et al. 1995a], Journal of Mediterranean Archaeology 8(1): 54-8.

PERNICKA, E. 1992. Evaluating lead isotope data [comments on Sayre et al. 1992a], Archaeometry 34(2): 322-6.

1993. Comments [Budd et al. 1993a], Archaeometry 35(2): 259-62.

1995. Crisis or catharsis in lead isotope analysis? [Comments on Budd et al. 1995a], Journal of Mediterranean Archaeology 8(1): 59-64.

PERNICKA, E., T.C. SEELIGER, G.A. WAGNER, F. BEGEMANN, S. SCHMITT-STRECKER, C. EIBNER, O. OZTUNALI & I. BARANYI. 1984. Archaometallurgische Untersuchungen in Nordwestanatolien, Jarbuch des Romisch-Germanischen Zentralmuseums, Mainz 31: 533-99.

PITTIONI, R. 1957. Urzeitlicher Bergbau auf Kupfererz und Spurenanalyse, Archaeologia Austriaca. Beiheft 1.

REEDY, T.J. & C.L. REEDY. 1992. Evaluating lead isotope data [comments on Sayre et al. 1992a], Archaeometry 34(2): 327-9.

SAYRE, E.V., K.A. YENER, E.C. JOEL & I.L. BARNES. 1992a. Statistical evaluation of the presently accumulated lead isotope data from Anatolia and surrounding regions, Archaeometry 34(1): 73-106.

1992b. Evaluating lead isotope data [comments on Sayre et al. 1992a], Archaeometry 34(2): 330-36.

1993. Comments [on Budd et al. 1993a], Archaeometry 35(2): 247-52.

1995. Comments [on Budd et al. 1995a], Journal of Mediterranean Archaeology 8(1): 45-53.

SCAIFE, B., P. BUDD, J.G. MCDONNELL, A.M. POLLARD & R.G. THOMAS. In press. A reappraisal of statistical techniques used in lead isotope analysis, in Proceedings of the International Symposium on Archaeometry, Ankara, May 1994.

SHERRATT, A. 1994. Core, periphery and margin: perspectives on the Bronze Age, in C. Mathers & S. Stoddart (ed.), Development and decline in the Mediterranean Bronze Age: 335-45. Sheffield: Sheffield Academic Press. Sheffield Archaeological Monographs 8.

STOS-GALE, Z., N.H. GALE, J. HOUGHTON & R. SPEAKMAN. 1995. Lead isotope data from the Isotrace Laboratory, Oxford: Archaeometry data base 1, ores from the Western Mediterranean, Archaeometry 37(2): 407-15.

TAYLOR, T. 1992. The Gundestrup Cauldron, Scientific American 266(3): 84-9.

THOMPSON, F.C. 1958. The early metallurgy of copper and bronze, Man 58: 1-6.